libsvm寻优matlab,libsvm参数寻优的准确率达不到

%重新分配训练和预测的样本

clc

close all

clear

num=[100,101,214,118,103,109,212,232];

len=length(num);

path='E:\dataandpro\Arrhythmia\';

featuretrain=[];

featurepre=[];

typetrain=[];

typepre=[];

wavename='db2';

feature=[];

featuretotal=[];

typetotal=[];

for i=1:len

number=num(i);

path1=strcat(path,num2str(number));

path2=strcat(path1,'\r_r_interval.mat');

load(path2);

% r_r_interval=[];

% for i=1:length(Rindnew)-1

% r_r_interval(i)=(Rindnew(i+1)-Rindnew(i))/360;

% end

path3=strcat(path1,'\s_t_interval.mat');

load(path3);

path4=strcat(path1,'\q_t_interval.mat');

load(path4);

path5=strcat(path1,'\type.mat');

load(path5);

path6=strcat(path1,'\XC.mat');

load(path6);

path7=strcat(path1,'\pstart_ind.mat');

load(path7);

% path8=strcat(path1,'\Rindnew.mat');

% load(path8);

a=[];

for j=2:length(pstart_ind)-1

[c,l]=wavedec(XC(pstart_ind(j):pstart_ind(j+1)),3,wavename);

a3=appcoef(c,l,wavename,3);

mina=min(a3);

maxa=max(a3);

meana=mean(a3);

stda=std(a3);

a=[a;mina,maxa,meana,stda];

end

avgR=mean(r_r_interval);

pavg=[];

pavg=r_r_interval./avgR;%每个RR间期除以平均值，可以用来提取Pre/svgR,post/avrg

s_t_interval=s_t_interval(2:length(s_t_interval));

q_t_interval=q_t_interval(2:length(q_t_interval));

len1=min(length(pavg),length(s_t_interval));

len2=min(length(type),length(q_t_interval));

len3=min(len1,len2);

[len4,wide]=size(a);

len=min(len3,len4);

%feature=[];

pavg=pavg';

s_t_interval=s_t_interval';

q_t_interval=q_t_interval';

feature=[pavg(1:len-1),pavg(2:len)];%,s_t_interval(1:len-1),q_t_interval(1:len-1)];

[len1,wide1]=size(feature);

[len2,wide2]=size(type);

len=min(len1,len2);

feature=[feature(1:len-1,:),a(1:len-1,:),a(2:len,:)];

type=type(1:len1-1);

featuretotal=[featuretotal;feature];

typetotal=[typetotal;type];

end

id=[1 2 3 5 8 13 ];

position1=[];

position2=[];

position3=[];

position5=[];

position8=[];

%position13=[];

%position_1=[];

len=length(typetotal);

for i=1:len

if typetotal(i)==1

position1=[position1;i];

elseif typetotal(i)==2

position2=[position2;i];

elseif typetotal(i)==3

position3=[position3;i];

elseif typetotal(i)==5

position5=[position5;i];

else typetotal(i)==8

position8=[position8;i];

end

end

len1=length(position1);

len2=length(position2);

len3=length(position3);

len5=length(position5);

len8=length(position8);

featuretrain=[featuretrain;featuretotal(position1(1:round(2/3*len1)),:)];

featuretrain=[featuretrain;featuretotal(position2(1:round(2/3*len2)),:)];

featuretrain=[featuretrain;featuretotal(position3(1:round(2/3*len3)),:)];

featuretrain=[featuretrain;featuretotal(position5(1:round(2/3*len5)),:)];

featuretrain=[featuretrain;featuretotal(position8(1:round(2/3*len8)),:)];

typetrain=[typetrain;typetotal(position1(1:round(2/3*len1)))];

typetrain=[typetrain;typetotal(position2(1:round(2/3*len2)))];

typetrain=[typetrain;typetotal(position3(1:round(2/3*len3)))];

typetrain=[typetrain;typetotal(position5(1:round(2/3*len5)))];

typetrain=[typetrain;typetotal(position8(1:round(2/3*len8)))];

featurepre=[featurepre;featuretotal(position1(round(2/3*len1)+1:len1),:)];

featurepre=[featurepre;featuretotal(position2(round(2/3*len2)+1:len2),:)];

featurepre=[featurepre;featuretotal(position3(round(2/3*len3)+1:len3),:)];

featurepre=[featurepre;featuretotal(position5(round(2/3*len5)+1:len5),:)];

featurepre=[featurepre;featuretotal(position8(round(2/3*len8)+1:len8),:)];

typepre=[typepre;typetotal(position1(round(2/3*len1)+1:len1))];

typepre=[typepre;typetotal(position2(round(2/3*len2)+1:len2))];

typepre=[typepre;typetotal(position3(round(2/3*len3)+1:len3))];

typepre=[typepre;typetotal(position5(round(2/3*len5)+1:len5))];

typepre=[typepre;typetotal(position8(round(2/3*len8)+1:len8))];

[bestacc,bestc,bestg]=SVMcgForClass(typetrain,featuretrain,-5,5);

% bestc=13.9288;寻得的最优参数

% bestg=0.0625';

% bestacc=96.6823

model=svmtrain(typetrain,featuretrain,'-s 0 -t 2 -c 13.9288 -g 0.0625');

[label,accrury,dec]=svmpredict(typepre,featurepre,model);

CR = ClassResult(typepre,featurepre,model,2);